Calibrating MPU6050

Question

I using a MPU 6050 and raspberry pi pico for making an IMU. I have set the range of accelerometer as +- 2 g but there is a .96 g offset this severly limits the sensing range on the positive side. I have written a calibration program which works fine in which we explicitly add the offsets to the data `

def calibrate(i2c):
    temp=0
    accel_x,accel_y,accel_z,gyro_x,gyro_y,gyro_z=0,0,0,0,0,0

    for i in range(10000):
        temp += read_raw_data(i2c, TEMP_OUT_H) / 340.0 + 36.53
        accel_x += read_raw_data(i2c, ACCEL_XOUT_H) / 16384.0
        accel_y += read_raw_data(i2c, ACCEL_XOUT_H + 2) / 16384.0
        accel_z += read_raw_data(i2c, ACCEL_XOUT_H + 4) / 16384.0
        gyro_x += read_raw_data(i2c, GYRO_XOUT_H) / 131.0
        gyro_y += read_raw_data(i2c, GYRO_XOUT_H + 2) / 131.0
        gyro_z += read_raw_data(i2c, GYRO_XOUT_H + 4) / 131.0
    
    print('calibration done')
    
    t=temp/10000
    a_x=accel_x/10000
    a_y=accel_y/10000
    a_z=accel_z/10000
    g_x=gyro_x/10000
    g_y=gyro_y/10000
    g_z=gyro_z/10000

is there any way in which i can calibrate the sensor and still get full range?

[Calibrating Accelerometer and Gyroscope](https://makersportal.com/blog/calibration-of-an-inertial-measurement-unit-imu-with-raspberry-pi-part-ii) - it's for the 9250, but the 6050 is exactly the same (same registers). The 6050 just lacks the magnetometer. Primary reference (aside from the datasheet) is the [MPU-6000 and MPU-6050 Register Map and Descriptions](https://invensense.tdk.com/wp-content/uploads/2015/02/MPU-6000-Register-Map1.pdf) — David C. Rankin, Jun 17 '23 at 05:42
Calibration is a 2-step process. First you need to read the factory self-test values for base offset values for each access as measured by the factory. (all MPUs very slightly just do to the manufacturing process). The self-test values store that information. Next you need to enable the FIFO and fill it while moving the MPU in a figure-8. From those values you can compute the accel & gyro scale (range) adjustments needed to fine tune. The values are also noisy and will bounce around zero. You can "de-noise" your outputs by ignoring changes of several hundredths on each axis. — David C. Rankin, Jun 17 '23 at 06:06
And yes, `1G` (or `0.96G` in your case, showing `0.04G` error on the axis) is just the MPU measuring gravity. Never goes away. (unless in orbit, etc..). You can drop a rock to confirm. `:)` Questions like this are better asked on [Electronics StackExchange](https://electronics.stackexchange.com/) — David C. Rankin, Jun 17 '23 at 06:13

score 2 · Accepted Answer · answered Jun 17 '23 at 03:41

The 0.96 g is gravitational acceleration and must be from the vertical axis (typically z-axis). This is not a sensor bias. Included in the accelerometer acceleration measurement is gravitational acceleration. And so the answer is no. Gravity will take up part of the measurement range.

To get a feel for typical accelerometer readings check out a sample dataset from https://imuengine.io/resources/. Copied here are the first few IMU entries from the automotive dataset where the z-axis nominally points down. Convert the units from m/s^2 to g's and you see the same approximate 1 g from gravity.

TimeFromStart (s)  AccelX (m/s^2)         AccelY (m/s^2)       AccelZ (m/s^2)      AngleRateX (rad/s)     AngleRateY (rad/s)      AngleRateZ (rad/s)
49.27              0.0065134831           0.1535978052         -9.8256081343       0.0033288721           0.0021099924            -0.000368629
49.28              -0.1163793611          0.1524602878         -9.7294434905       -0.0014737246          -0.001269474            -5.3143600000000004e-05
49.29              -0.1772884163          0.1816529664         -9.833201021        -0.0009057809          -0.0002318527           -0.0004045664

Measuring "acceleration minus gravity" is also possible but requires a layer of sensor fusion. The method shown in the original post falls apart if the sensor is tilted. Feel free to post a new question detailing the requirements if that is what you need then.

May also be worth noting if it is a "How does this part of the MPU work..." verses an actual coding problem, then it is better to migrate the question to [Electronics StackExchange](https://electronics.stackexchange.com/). (explaining what is being measured isn't really a coding problem, but good effort in helping with the question) — David C. Rankin, Jun 17 '23 at 06:09

Calibrating MPU6050

1 Answers1